[PATCH] mm: reduce spinlock contention in release_pages()

[Hao Lee <haolee.swjtu@xxxxxxxxx>]

When several tasks are terminated simultaneously, lots of pages will be
released, which can cause severe spinlock contention. Other tasks which
are running on the same core will be seriously affected. We can yield
cpu to fix this problem.

Signed-off-by: Hao Lee <haolee.swjtu@xxxxxxxxx>
---
 include/linux/memcontrol.h | 26 ++++++++++++++++++++++++++
 mm/memcontrol.c            | 12 ++++++++++++
 mm/swap.c                  |  8 +++++++-
 3 files changed, 45 insertions(+), 1 deletion(-)

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index 0c5c403f4be6..b06a5bcfd8f6 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -800,6 +800,8 @@ struct lruvec *folio_lruvec_lock(struct folio *folio);
 struct lruvec *folio_lruvec_lock_irq(struct folio *folio);
 struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
 						unsigned long *flags);
+struct lruvec *folio_lruvec_trylock_irqsave(struct folio *folio,
+						unsigned long *flags);
 
 #ifdef CONFIG_DEBUG_VM
 void lruvec_memcg_debug(struct lruvec *lruvec, struct folio *folio);
@@ -1313,6 +1315,17 @@ static inline struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
 	return &pgdat->__lruvec;
 }
 
+static inline struct lruvec *folio_lruvec_trylock_irqsave(struct folio *folio,
+		unsigned long *flagsp)
+{
+	struct pglist_data *pgdat = folio_pgdat(folio);
+
+	if (spin_trylock_irqsave(&pgdat->__lruvec.lru_lock, *flagsp))
+		return &pgdat->__lruvec;
+	else
+		return NULL;
+}
+
 static inline struct mem_cgroup *
 mem_cgroup_iter(struct mem_cgroup *root,
 		struct mem_cgroup *prev,
@@ -1598,6 +1611,19 @@ static inline struct lruvec *folio_lruvec_relock_irqsave(struct folio *folio,
 	return folio_lruvec_lock_irqsave(folio, flags);
 }
 
+static inline struct lruvec *folio_lruvec_tryrelock_irqsave(struct folio *folio,
+		struct lruvec *locked_lruvec, unsigned long *flags)
+{
+	if (locked_lruvec) {
+		if (folio_matches_lruvec(folio, locked_lruvec))
+			return locked_lruvec;
+
+		unlock_page_lruvec_irqrestore(locked_lruvec, *flags);
+	}
+
+	return folio_lruvec_trylock_irqsave(folio, flags);
+}
+
 #ifdef CONFIG_CGROUP_WRITEBACK
 
 struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb);
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 781605e92015..b60ba54e2337 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1261,6 +1261,18 @@ struct lruvec *folio_lruvec_lock_irqsave(struct folio *folio,
 	return lruvec;
 }
 
+struct lruvec *folio_lruvec_trylock_irqsave(struct folio *folio,
+		unsigned long *flags)
+{
+	struct lruvec *lruvec = folio_lruvec(folio);
+
+	if (spin_trylock_irqsave(&lruvec->lru_lock, *flags)) {
+		lruvec_memcg_debug(lruvec, folio);
+		return lruvec;
+	}
+
+	return NULL;
+}
 /**
  * mem_cgroup_update_lru_size - account for adding or removing an lru page
  * @lruvec: mem_cgroup per zone lru vector
diff --git a/mm/swap.c b/mm/swap.c
index e8c9dc6d0377..91850d51a5a5 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -960,8 +960,14 @@ void release_pages(struct page **pages, int nr)
 		if (PageLRU(page)) {
 			struct lruvec *prev_lruvec = lruvec;
 
-			lruvec = folio_lruvec_relock_irqsave(folio, lruvec,
+retry:
+			lruvec = folio_lruvec_tryrelock_irqsave(folio, lruvec,
 									&flags);
+			if (!lruvec) {
+				cond_resched();
+				goto retry;
+			}
+
 			if (prev_lruvec != lruvec)
 				lock_batch = 0;
 
-- 
2.31.1